⚙️ OpenDroneMap Infrastructure Projects 🛠️

[smathermather]

Intro

One of the goals of OpenDroneMap having the mantle of an org is ensuring we support appropriate infrastructure for the ecosystem. I thought I would start a thread on things we could be doing to this end. This will likely serve as a starting place for a more formal list, but I thought it might be a bit more accessible to start it here rather than a list on github or similar.

What do I mean by infrastructure? Some of this is stuff already happening. We

• UAVArena:
  • Let's compare ODM! [Data Challenge, earn a badge]
  • UAV Mapping Arena
• ODMData
  • GitHub - OpenDroneMap/ODMdata: Example data for OpenDroneMap: https://OpenDroneMap.org sourced from a variety of sources
• A bunch of build structures on github and elsewhere
• This community forum!
• OpenDroneMap.org main site

And a bunch of things other things. I’m being bad and not cataloging all of them now… .

The Call

Let’s generate a list of infrastructure, both what we have (like the above) and what we would like to have (including refinements to what we have.

[smathermather]

I’ll add some things that I’d like to see us do:

• A tool that solves performs analyzes drone imagery for pitch/roll/yaw (or your favorite angular system) to allow for super fast triangulation
  • This tool would take the results from an incremental reconstruction and derive a reference make, model, firmware version, and a formula or map between EXIF/XMP tags and calculated angles
  • It could be:
    • a hosted piece of infrastructure with some sort of database of known values
    • or just some code you can run on your own data
    • or both with submissions to hosted infrastructure either being raw data or processed results
  • Some chatter about it here: Differences between Triangulation and Incremental SfM - #8 by smathermather
  • And reference for OpenSfM angles here: Rotation Angles - #2 by pierotofy
  • ODM/photo.py at master · OpenDroneMap/ODM · GitHub
• More formal curation of test / reference datasets. Piero and others do a fantastic job testing algorithms as they change, but having a suite of test datasets cataloged for referencing changes in data quality, performance, etc. would be super useful.
• Test infrastructure for building data products from reference datasets, so with each commit or tag or (insert threshold here), we have a publicly hosted set of updated reference datasets and their respective performance metrics
• A log reader / catalog of performance tests across hardware for estimating completion times based on image size, number of images, processing settings, processor type, memory, etc. etc… Similar to the angle tool above, it could be both a piece of code and an infrastructure with curated dataset for better understanding
• Curation of point cloud datasets for e.g. classification per Piero’s call here Point Cloud Classification Datasets?

Ok, there’s a bunch more that pops into my brain and then pops back out, and maybe yours too. Put them here if they do.

[pierotofy]

• Issues · OpenDroneMap/oats · GitHub and GitHub - OpenDroneMap/oats: OpenDroneMap Automated Testing Suite

[smathermather]

Aha, yes indeed!